Gespeichert in:
Bibliographische Detailangaben
Hauptverfasser: Norelli, Antonio, Bronstein, Michael
Format: Preprint
Veröffentlicht: 2025
Schlagworte:
Online-Zugang:https://arxiv.org/abs/2510.20075
Tags: Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
_version_ 1866911379932315648
author Norelli, Antonio
Bronstein, Michael
author_facet Norelli, Antonio
Bronstein, Michael
contents A meaningful text can be hidden inside another, completely different yet still coherent and plausible, text of the same length. For example, a tweet containing a harsh political critique could be embedded in a tweet that celebrates the same political leader, or an ordinary product review could conceal a secret manuscript. This uncanny state of affairs is now possible thanks to Large Language Models, and in this paper we present Calgacus, a simple and efficient protocol to achieve it. We show that even modest 8-billion-parameter open-source LLMs are sufficient to obtain high-quality results, and a message as long as this abstract can be encoded and decoded locally on a laptop in seconds. The existence of such a protocol demonstrates a radical decoupling of text from authorial intent, further eroding trust in written communication, already shaken by the rise of LLM chatbots. We illustrate this with a concrete scenario: a company could covertly deploy an unfiltered LLM by encoding its answers within the compliant responses of a safe model. This possibility raises urgent questions for AI safety and challenges our understanding of what it means for a Large Language Model to know something.
format Preprint
id arxiv_https___arxiv_org_abs_2510_20075
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle LLMs can hide text in other text of the same length
Norelli, Antonio
Bronstein, Michael
Artificial Intelligence
Computation and Language
Cryptography and Security
Machine Learning
A meaningful text can be hidden inside another, completely different yet still coherent and plausible, text of the same length. For example, a tweet containing a harsh political critique could be embedded in a tweet that celebrates the same political leader, or an ordinary product review could conceal a secret manuscript. This uncanny state of affairs is now possible thanks to Large Language Models, and in this paper we present Calgacus, a simple and efficient protocol to achieve it. We show that even modest 8-billion-parameter open-source LLMs are sufficient to obtain high-quality results, and a message as long as this abstract can be encoded and decoded locally on a laptop in seconds. The existence of such a protocol demonstrates a radical decoupling of text from authorial intent, further eroding trust in written communication, already shaken by the rise of LLM chatbots. We illustrate this with a concrete scenario: a company could covertly deploy an unfiltered LLM by encoding its answers within the compliant responses of a safe model. This possibility raises urgent questions for AI safety and challenges our understanding of what it means for a Large Language Model to know something.
title LLMs can hide text in other text of the same length
topic Artificial Intelligence
Computation and Language
Cryptography and Security
Machine Learning
url https://arxiv.org/abs/2510.20075